Navigating the Dual Facets: A Comprehensive Evaluation of Sequential Memory Editing in Large Language Models

arXiv:2402.11122v1 Announce Type: new
Abstract: Memory Editing (ME) has emerged as an efficient method to modify erroneous facts or inject new facts into Large Language Models (LLMs). Two mainstream ME methods exist: parameter-modifying ME and parameter-preserving ME (integrating extra modules while preserving original parameters). Regrettably, previous studies on ME evaluation have two critical limitations: (i) evaluating LLMs with single edit only, neglecting the need for continuous editing, and (ii) evaluations focusing solely on basic factual triples, overlooking broader LLM capabilities …

abstract arxiv cs.ai cs.cl editing evaluation extra facts language language models large language large language models llms memory modules parameters type